How to choose an AI automation partner
You've been told you should be "doing something with AI." Maybe a business advisor said so. Maybe your team is spending half of Monday on a process that shouldn't take an hour. Maybe you've simply read enough headlines to feel you're being left behind.
The trouble is that the moment you start looking, everyone is an AI expert. And for practical, small-business AI automation, the field is still young — which makes the usual way of choosing a supplier, a long and proven track record, harder to lean on. Most providers have a shorter track record with the current generation of AI tools than they do with technology more broadly.
So this is a guide to choosing well anyway. It's written by a company that does this work for a living — so we're not a neutral party — but the questions below are fair ones, and they're worth putting to anyone. Including us.
Who's actually offering this
"AI automation" is being sold by very different kinds of business. It helps to know which one you're talking to.
The marketing or "AI" agency
Strong on communication, positioning and the user-facing story, and often genuinely excellent at it. Positioning and building are different jobs, and the build is sometimes subcontracted. So ask plainly who is doing the technical work, and who is responsible once it is live.
The large software house
Genuinely capable of building almost anything, and organised around larger clients. The cost, the pace and the layers of account management can be a poor fit for a small business, so ask how a small account is looked after.
The marketplace freelancer
Fiverr, Upwork, PeoplePerHour. There are capable people there, and the price is low. The trade-off is that you may need to take on more of the project-management and systems-integration role yourself. It works well if you already know exactly what you want and how it should be built, and simply need hands to build it.
The small developer-led firm
Small enough to deal with you directly, but with the technical depth to design, build, deploy and support the whole thing themselves.
Full disclosure: that last category is ours. But the point of this page isn't the sales pitch — it's the questions below, and they're fair whoever you put them to.
The important thing isn't what a provider calls itself. It's whether they can design it, build it, connect it to your systems, secure it, and support it afterwards.
The questions worth asking anyone
Put these to anyone you're considering — including us.
1. Who will actually build it?
Ask plainly: will the people in the room write this, or will it be subcontracted? It matters for cost, and for how well the finished thing fits — but most of all for what happens later. When something needs changing in eight months' time, you want to know who will still understand the system and be responsible for helping you change it.
2. What have they actually shipped?
Nobody has ten years of experience with the current generation of AI automation tools, so treat very long track-record claims with care. But two or three years of real, working delivery is genuinely possible. So ask to see it. What have they built in the last couple of years that is running in a real business right now? Can you look at it? Can you speak to the client? A short track record is fine. No track record, dressed up as confidence, is not.
3. Will it run on your actual systems?
This one is easy to miss. Whatever gets built has to run somewhere — and it has to connect to the systems you already use: your email, your files, your existing software, your data. That makes it, underneath, an IT job as much as an AI one. A provider who understands the infrastructure it will live on can deploy it properly, secure it, and fit it to how your business already works. A provider who only does "the AI" can hand you something clever that never quite fits the way your business actually works.
4. Who is responsible once it's live?
Automation isn't fit-and-forget. The AI tools themselves change; your business changes; things break. Ask who is still responsible six months on. A firm that builds and supports its own work can stay responsible for it after launch. If delivery was subcontracted, ask clearly who will support the finished system.
5. Where does your data go?
AI automation usually means your business's information passing through a model or a third-party service — so it's worth asking exactly what happens to it. Which model or service is being used? Where is the data processed, and is any of it retained? Is it used to train someone else's system? Who can see it? You don't need to be technical to ask — but you do want a provider who can answer clearly, in plain English, and explain the GDPR position without hand-waving.
6. Will they tell you when not to do it?
A good provider should be willing to say when AI is not the right answer. Sometimes a process isn't worth automating. Sometimes the honest answer is a £20-a-month off-the-shelf tool, not a bespoke build. A partner willing to talk you out of work is showing you exactly the judgement you're paying for.
7. Are you buying advice, implementation, or both?
Some firms sell strategy: a readiness assessment, methodology or roadmap. That can be useful, particularly when the problem is not yet well defined. What matters is knowing what you are buying and what happens next.
Is the output a document? A prioritised plan? A working system? Will the same people who advised you be able to build it?
A methodology is not a result — but good strategy can be the thing that gets you to the right result.
What a good answer sounds like
It's one thing to know the questions. It helps to know what a solid answer sounds like — so here, roughly, is what you should hope to hear:
"We'd build this ourselves, not subcontract it. It would run against your existing Microsoft 365 setup and connect to the systems you already use. Your data wouldn't be used to train public models. We'd document what it does, support it after launch — and we'd start with one small process to prove it works before going any wider."
You don't need every word of that. But if a provider can talk in those terms — concrete, careful about scope, clear about ownership and about data — you are talking to someone who has done this before. If the answers stay vague, or every reply is an enthusiastic yes, that tells you something too.
Choosing well matters more than choosing us
We wrote this because a badly chosen AI project wastes money and sours people on the whole idea — and we would genuinely rather you chose well than chose us blindly. If these questions lead you to us, we'd be glad to talk. If they lead you somewhere else, that's a good outcome too. Either way, ask them.
And if you'd like to talk it through with someone who builds this for a living — no pitch, no jargon — the first conversation is free.
01684 899403